Bipolar Junction Transistor and CPU

Click For Summary
SUMMARY

The discussion centers on the role of Bipolar Junction Transistors (BJTs) in CPU design, specifically addressing the transition to CMOS technology. Participants clarify that while BJTs can theoretically create logic gates, their power consumption makes them impractical for modern CPUs, which utilize millions of transistors in integrated circuits. The conversation also highlights the importance of understanding voltage levels (Vcc, Vin, Vout) and the intricate processes involved in fabricating silicon wafers into functional CPUs. Historical references to early computers like the IBM 360 and Honeywell Sigma 8 illustrate the evolution of transistor technology in computing.

PREREQUISITES
  • Understanding of Bipolar Junction Transistors (BJTs) and their operation
  • Knowledge of CMOS technology and its advantages over BJTs
  • Familiarity with integrated circuit design and logic gates
  • Basic concepts of semiconductor physics and voltage levels (Vcc, Vin, Vout)
NEXT STEPS
  • Research the differences between Bipolar Junction Transistors and CMOS technology
  • Learn about integrated circuit design principles and logic gate implementation
  • Explore semiconductor fabrication processes, particularly photolithography
  • Investigate the historical development of early computers and their transistor technologies
USEFUL FOR

Electrical engineers, computer architects, and anyone interested in the evolution of transistor technology in computing and integrated circuit design.

fredreload
Messages
250
Reaction score
6
I am looking at the Bipolar junction transistor, but I cannot seem to correlate with how the design would work for the CPU in terms of 0 and 1s. I am looking mainly at the transistor diagram from Wikipeida https://en.wikipedia.org/wiki/Transistor. Can someone tell me what Vcc, Vin, and Vout are as well as how they are incorporated into the CPU design. You have voltage input from base which controls the flow of a larger current from collector to emitter thanks to the Semiconductor Physics but you would have to manually control Vin for the base supply voltage to turn the larger current on or off as 0 and 1s. Do you use logic gates to control these transistors? Why would modern electronic device needs a few million of these transistors and how are they paired, in a tree structure? I know my questions seem kind of vague but if someone can enlighten me with the answer it would be cool. After all a few million of these transistors fit into a computer in the present day.
 
Engineering news on Phys.org
CPUs don't use bipolar transistors, they use CMOS.

It is possible to make the standard logic components out of bipolar transistors (e.g. AND), so a BJT-based CPU would be possible in theory. It would require constant current in all those logic elements, however, so power consumption would be completely impractical.
 
A long time ago, they did use discrete bipolar transistors for computers. For example, the famous IBM 360.

Here is an article http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=5215804&url=http%3A%2F%2Fieeexplore.ieee.org%2Fiel5%2F6%2F5215778%2F05215804.pdf%3Farnumber%3D5215804, from 1967.

In the 1970's I worked with a bipolar Honeywell Sigma 8 computer. It needed 75 tons of air conditioning.
 
BJT's (or CMOS FETs) can be used as individual switching devices (on or off). You need to understand how that can be used to do simple logic functions.

These switches are used as the basic components in logic gates (https://en.wikipedia.org/wiki/Logic_gate)

Millions of logic gates are combined to design "Integrated circuits". ( https://en.wikipedia.org/wiki/Integrated_circuit).

Dozens to hundreds of integrated circuits are combined to build computer components (motherboards, power supplies, hard drives, etc).
The processor architecture is one part of the system https://en.wikipedia.org/wiki/Computer_architecture

Your PC power supply supplies 3.3V and 12V which is used by various power control systems in the computer to supply the voltages needed by the individual transistors.

It is a long road from transistor to computer.
 
You can still buy BJT logic devices but their corner is pretty narrow today. You can still see it (ECL) used in applications that need low noise logic like cryptographic devices with a requirement for very low operational signal emissions to reduce side-channel attacks and It's still a common interface in high speed video networks as PECL, etc... http://pdfserv.maximintegrated.com/en/an/AN291.pdf
 
  • Like
Likes Jeff Rosenbury

I was looking at this video in how Global Foundaries turn a Silicon wafer into CPU. At around 3:50 it is explained that the circuit is etched onto the wafer with a photo lithographic process and tiny transistors are formed, but it doesn't really show how it works at an atomic level. If someone has more information about how Silicon wafer works as a transistor on an atomic level after the doping process, 5:32, and after the copper is wired on Silicon disk (7:00), it would be cool. From what I can see the circuit and copper wire is all layered on top of the original Silicon wafer, that means when the current is running in the copper wire it would run through these Silicon transistors? How do these billions of transistors work together?
Here's the Intel one, it seems to have more steps than Global Foundaries, but without explanations.

P.S. I found the answer, sorry for all the excessive video posts.
 
Last edited:
Nevermind I still can't find how CPU utilize every single transistor on the board. I mean the integrated circuit imprinted onto the wafer is bigger than a transistor right? Do they actually shape like it is shown in the Youtube video below?



P.S. Wait I really found it this time, the integrated circuit is built on top of the transistors, I thought it's burned onto it with the lithographic process.
 
Last edited:
The integrated circuit consists of millions of transistors connected by wires. It is the combination of the transistors and how they are connected.

You could (and some old gate arrays did this) start with a circuit that consisted of an array of transistors. You could then put down a layer of metal and etch out what you don't want to give you a circuit consisting of transistors connected by metal "wires", analogous to a PCB.

The process starts with a silicon wafer. Many many things happen (hundreds of steps), one after the other, to leave you with millions of transistors connected by wires.
Connecting all those transistors in a way that produces meaningful functional behavior is a MASSIVE expansion of that AND function video.
 
meBigGuy said:
The integrated circuit consists of millions of transistors connected by wires. It is the combination of the transistors and how they are connected.

You could (and some old gate arrays did this) start with a circuit that consisted of an array of transistors. You could then put down a layer of metal and etch out what you don't want to give you a circuit consisting of transistors connected by metal "wires", analogous to a PCB.

The process starts with a silicon wafer. Many many things happen (hundreds of steps), one after the other, to leave you with millions of transistors connected by wires.
Connecting all those transistors in a way that produces meaningful functional behavior is a MASSIVE expansion of that AND function video.
Well, not "wires" in the conventional sense. More like conducting channels which are part of the integrated circuit.

https://en.wikipedia.org/wiki/Integrated_circuit_design
 
  • #10
SteamKing said:
Well, not "wires" in the conventional sense. More like conducting channels which are part of the integrated circuit.

Wires in the same way that you can consider PCB traces as wire.
Do you call the innerlayer traces in a PCB "channels"?
Copper or aluminum is deposited and etched and vias are created to wire together the components. Then a layer of insulator is applied, another layer of metal, and more etching (ignoring the steps for via formation)

When I read a summary report for our small 3mm x 3mm 180nm IC's I see 14 meters of "wire" (which blows my mind).

If you choose to not call metal based interconnect "wire" that is your choice. The industry calls it "wire".

Do a search on "ASIC wire length"

http://asic-soc.blogspot.com/2008/10/net-delay.html for example

I have designed ASIC's for a living for many years, and I am still amazed by what it represents. Take this slab of crystallized glass, expose it to chemicals and deposit stuff on it and when you get it back it does almost exactly what the simulators said it would do. Like I said, 9mm^2 with 14 meters of wire (one old 180nm mixed signal example I remembered the numbers for).

The lineup of tools and technology to make that happens is beyond comprehension, from the semiconductor process designers, to the epi reactor and metal deposition machine designers, to the simulation and layout tools, to the functional emulation and verification systems, to the testing and verification flows, and on and on and on.
Everytime an IC comes back and the processor boots first time just like it did on the FPGA based emulation system and data flows as it should, I get this sense of seeing Magic.
 
  • Like
Likes mfb
  • #11
meBigGuy said:
Wires in the same way that you can consider PCB traces as wire.
Do you call the innerlayer traces in a PCB "channels"?
Copper or aluminum is deposited and etched and vias are created to wire together the components. Then a layer of insulator is applied, another layer of metal, and more etching (ignoring the steps for via formation)

When I read a summary report for our small 3mm x 3mm 180nm IC's I see 14 meters of "wire" (which blows my mind).

If you choose to not call metal based interconnect "wire" that is your choice. The industry calls it "wire".
Yes, the industry may call it "wire", but not everyone at PF is in the industry.

If this were ICforum, you would have a point, but there are a lot of people who use PF who are not technical types and may not understand all the nuances of industry jargon.
 
  • #12
SteamKing said:
If this were ICforum, you would have a point, but there are a lot of people who use PF who are not technical types and may not understand all the nuances of industry jargon.

So, let's develop some ambiguous and incorrect jargon of our own to keep them totally confused.

I think PCB traces present a simple and accurate analogy. A technically neutral term would be "metal interconnect". I had a serious issue with the connotations of "conducting channel" which sort of implies other things like something that was dug and filled, or the conducting diffusion channel in a FET, or polysilicon. Also, in an IC, "routing channels" specifically means regions of the chip set aside for groups of metal interconnect.

Do these, or the pictures in post #6 look like "conducting channels"? ( from https://en.wikipedia.org/wiki/Integrated_circuit which BTW, does not even contain the word "channel")
220px-Silicon_chip_3d.png


PS, your link in the post where you present your concept of "conducting channels" also does not contain the words "channel" or "channels"
 

Similar threads

  • · Replies 23 ·
Replies
23
Views
4K
Replies
80
Views
5K
  • · Replies 5 ·
Replies
5
Views
3K
  • · Replies 3 ·
Replies
3
Views
1K
Replies
68
Views
7K
  • · Replies 5 ·
Replies
5
Views
4K
  • · Replies 11 ·
Replies
11
Views
2K
Replies
2
Views
2K
Replies
10
Views
2K
  • · Replies 8 ·
Replies
8
Views
2K